Archetypes of Supply Chain Analytics Initiatives — An Exploratory Study

While Big Data and Analytics are arguably rising stars of competitive advantage, their application is often presented and investigated as an overall approach. A plethora of methods and technologies combined with a variety of objectives creates a barrier for managers to decide how to act, while researchers investigating the impact of Analytics oftentimes neglect this complexity when generalizing their results. Based on a cluster analysis applied to 46 case studies of Supply Chain Analytics (SCA) we propose 6 archetypes of initiatives in SCA to provide orientation for managers as means to overcome barriers and build competitive advantage. Further, the derived archetypes present a distinction of SCA for researchers seeking to investigate the effects of SCA on organizational performance.


Introduction
Even before data got their mainstream reputation of being the "new oil" of the 21st century predestined to shape the digital economy [1], the potential competitive advantages through data analytics had already been recognized [2].To remain with this analogy, data analytics represents the refinery process turning raw data into competitive strength.Analytics in itself is not a leading-edge invention, but increased attention was recently triggered by developments in information technology (IT) providing new access to data, organizations' need for better and faster decision making and recent big data and machine learning tools enabling new levels of insight for decision makers [3].The field of Logistics and Supply Chain Management (LSCM), has been identified as one early adopter and a long-term user of Analytics [4].LSCM has been employing operations research approaches for decades und uses purpose-specific analytics tools for very particular problems.As it is concerned with effectively integrating suppliers, manufacturers, warehouses, and stores, such that merchandise will be produced and distributed to the customer with a satisfying service level and with minimal costs in the right manner concerning time, location and quantity [5], the necessity for analytical approaches to achieve these efficiency goals is inevitable.A recent industry report underlines the long-term data affinity of the LSCM sector from a practical point of view and the potential of logistics operations generating the amounts of data needed to create value by using Analytics [6].However, the report emphasizes untapped potential in improving operational efficiency, customer experience or creating new business models in the sector.The fit of LSCM and Analytics is favorable since the satisfaction of customer needs and requirements is a leading theme in LSCM ( [7] p. 12) and the meaningful insight provided by Analytics is used to ensure rules and workflows strengthening satisfaction of needs and requirements [8].
Scholars and practitioners alike have provided evidence that advantages to performance can be achieved from the domain specific use of Analytics, which was termed "Supply Chain Analytics" [9,10].In research, several authors have investigated the effects of Supply Chain Analytics.Information system supported Analytics capabilities have been indicated to improve LSCM performance [11] and Analytics tends to have a positive impact on LSCM performance with high dependency on the fit of Analytics investment and LSCM process maturity [12].The positive effect of Analytics on LSCM was further suggested as context contingent, especially on planning processes [13].Schoenherr and Speier-Pero [14] provide a wide variety of perceived benefits including improved supply chain efficiency and decreased supply chain costs.Furthermore, several scholars call for more research on the topic [15,16].
On the practitioners' side, industry reports have shown high expectations towards Analytics in LSCM, especially for reducing inventory, risk and improving batch sizes.Among others, better customer service, higher efficiency and faster reaction to supply chain issues based on investments in Analytics have been reported [17].Furthermore, the development of more customer-oriented value chains with lower logistics costs due to the rise of data and Analytics has been suggested [18].However, investments are somewhat reluctant [19] and industry reports show that only a few firms achieve excellent performance in Analytics concerning LSCM [20].The majority of firms is lagging or struggling with Analytics in LSCM [21] with managers reporting missing experience and lack of knowledge of how to apply Analytics [22].
In the studies summarized above, Analytics is customarily considered as one overall concept while making conclusions on it or deriving potential value and utility although single examples with individual issues and diverse analytical techniques are considered.Especially single examples are used to highlight Analytics providing benefits [11] or savings [21] thereby projecting the exemplary benefits to Analytics as a general concept or overall approach.What remains unknown are the different outcomes of different approaches in relation to their intentions and execution.Analytics is usually not subdivided further although it presents a wide field with different objectives, orientations, and perspectives without a unified definition [9].In our view, it is too wide of a field to assume that all reported effects can equally be applied on different Supply Chain Analytics initiatives, all initiatives having the same potential of providing value for an organization or all barriers appearing could be overcome in a single manner.The sole attempt to subdivide Supply Chain Analytics in extant research can be found in the framework for Analytics applications in LSCM [23], which focuses on off-the-shelf IT Systems and therefore ignores important aspects of Analytics as well as initiatives which are not system-specific.To derive more sophisticated and reliable research conclusions on the effects of Analytics on LSCM, a distinction of Supply Chain Analytics approaches is needed.We propose a distinction of how organizations apply Supply Chain Analytics, by using clustering on 46 case studies on Supply Chain Analytics initiatives considering intended problem to be solved, execution, techniques, and the resulting Analytics solution.Thus, this research investigates patterns in the activities of organizations applying Analytics to business problems in LSCM and explores their endeavors and motivation to form archetypes of initiatives with exclusive characteristics.The outcome of the initiatives as well as alignment of outcome and intention is out of scope for research.Regarding MacInnis Framework of conceptual contributions, the goal of this study is differentiation [24].Thus, we will indicate how the identified archetypes are different, why this differentiation matters, and how they can be used further.The study is based on publications about manufacturing firms, retailers and logistics service providers applying Analytics in LSCM.The research questions therefore states: How can Supply Chain Analytics initiatives be distinguished?
The obtained archetypes can provide guidance to managers for their individual issues, and points of references of other organizations' previous activities, and thus, reduces barriers to adopt Supply Chain Analytics in their own organization.The archetypes represent types that are designed to be most different from each other to support learning of managers and students about Supply Chain Analytics.However, the combination of characteristics of different archetypes in the creation phase of a new initiative in an organization is not relegated but rather encouraged with an individual and specific goal and approach to be designed by the executing manager.For researchers, the archetypes form a framework to investigate the different effects of varying approaches.
The remainder of the article is structured as follows: Section 2 provides a theoretical background with the objective to explain the characteristics chosen to form the archetypes.Section 3 presents the methodology on how archetypes are formed using cluster analysis.Section 4 explains the suggested archetypes and discusses their impact.Section 5 concludes the article and Section 6 provides final remarks.

Theoretical Background
In this section, we will summarize Analytics, Supply Chain Analytics, and characteristics of Supply Chain Analytics initiatives.

Analytics
Due to its novelty and evolving nature, a wide variety of definitions of Analytics exists.Holsapple et al. ([9], p. 134) reviewed many of them to develop a collective definition stating that Analytics is "concerned with evidence-based problem recognition and solving that happen within the context of business situations".This definition highlights two specific aspects of Analytics.The first aspect, problem recognition, indicates the experimental part of Analytics to achieve a goal which is uncertain and unclear in the beginning requiring further exploration [25], and thus identifying what the actual problem is.The second aspect, problem solving, indicates that the value of Analytics is solely provided if a model or application is deployed and used [25].This aspect of Analytics is emphasized prominently in the literature, often specified as making decisions and taking actions (e.g., [2,8,26,27]).Both aspects establish a clear distinction from data aggregation initiatives such as dashboards and reports.
Davenport and Harris [2] presented the benefits of applying Analytics to improve internal processes or an organization's competitive position.They illustrated that achieving success with Analytics is not based on deploying software but rather on three categories of factors: organizational, human, and technological capabilities.Organizational capabilities consider analytical objectives and processes, human capabilities consider skills, sponsorship and culture, and technological capabilities consider data availability and Analytics architecture.While the models and software are often in the focus of research in Analytics due to apparent presentation of insight into the specific opportunities of Analytics, scholars highlight all three stated capabilities as critical to develop and successfully use models and software [8,28].
The models and software used in Analytics are commonly distinguished as being descriptive, predictive, or prescriptive (e.g., [9,23,29]).The meaning of descriptive analytics is twofold.On the one hand, it presents the summary of data to report and monitor [23].On the other hand, it describes root cause analysis used to gain insights about the underlying phenomenon or process [30,31].Predictive analytics estimates unknown values based on known examples.Prescriptive analytics determines and, in some cases, subsequently automates actions or decisions to achieve an objective given current and projected data, requirements and constraints.
Due to recent technological advances, Analytics gained additional interest as "Big Data Analytics", referring to Analytics performed with Big Data, which has been reported to have a positive impact on firm performance [32].Big Data originates in data management issues with technology in the early 2000s due to high volume, velocity, or variety of data [33], which formed the original three "V's" of Big Data.Big Data is momentarily under frequent academic investigation including an increase of "V's" (e.g., [32,34,35]) considering several issues with Big Data beyond the aspects of data management and without the need for advanced technologies such as distributed storage and processing, such as Variability, Veracity, Visualization or Value.However, three "V's" is a leading theme [14][15][16]26,28,30,36].

Supply Chain Analytics
Like Analytics, no unified definition of Supply Chain Analytics (SCA) exists, while rarely one is proposed.Souza ([10], p. 595) describes it as "focus[ing] on the use of information and analytical tools to make better decisions regarding material flows in the supply chain".Waller and Fawcett ([15], p. 79) propose a definition while describing the field as (L)SCM data science: "( . . . ) is the application of quantitative and qualitative methods from a variety of disciplines in combination with (L)SCM theory to solve relevant (L)SCM problems and predict outcomes, taking into account data quality and availability issues."Incorporating aspects of both descriptions and the definition of Analytics [9], we propose to define SCA as follows: SCA is concerned with evidence-based problem recognition and solving within the context of logistics and supply chain management situations.
Consequentially, SCA is neither a single and clear step-by-step approach to solve supply chain problems nor limited to certain tasks and processes in LSCM.Souza [10] systemizes and distinguishes several techniques by the type of Analytics and the SCOR processes affected.The origin of the list of techniques is not explained and it is not exhaustive.Furthermore, another attempt on systemization results for an investigation on in-memory technology used in LSCM by grouping in-memory software applications for LSCM and designing a framework for analytical applications [23].By considering the type of Analytics applied, whether the concept is data driven or model driven and methodological requirements, off-the-shelf software applications with analytical capabilities used for LSCM functionalities were grouped into monitor-and-navigate, sense-and-respond, predict-and-act, and plan-and-optimize.However, this categorization ignores objectives, organizational aspects, and human aspects.Finally, examples of potential applications of Analytics in LSCM were summarized from the perspectives of the user and the tasks [15].In summary, scholars have stretched a wide range of applications of SCA with various use cases for different functionalities and users, providing evidence that SCA is too complex to evaluate its impact as a general concept.
The generalization has further impact on managers by creating barriers, which we want to address with this study.Thus, this research focuses on barriers related to a missing understanding of how to apply SCA on individual problems of an organization and substantiate relevant SCA initiatives.Sanders [28] provides an extensive overview on barriers of Analytics in the context of LSCM and presents several barriers of which the following are related to the interest of this research.First, managers, especially in leadership positions, may not see the value provided by Analytics resulting in missing vision, understanding of the full capacity and how to change the organization to apply Analytics successfully.Second, so called analysis paralysis hinders organizations from applying Analytics because they cannot handle the overwhelming opportunities, the speed of technological change what results in the inability to define a starting point.Organizations may thus try to randomly analyze data for some eventual causation, some business units may optimize their sub processes with little global effect or organizations try to measure everything at once without understanding what to focus on.Third, instead of experiencing a lack of data, many organizations drown in data.Besides technological issues to handle these amounts of data, organizations do not know how to leverage the existing data capability and how to base decisions on it.

Dismantling Supply Chain Analytics Initiatives
This subsection describes the characteristics used to analyze SCA initiatives to form archetypes.We identified 34 characteristics in an extensive review of Analytics literature which are presented in six categories.The characteristics and categories are presented in Figure 1.Drawing on Chae et al. (2014) [13], we consider SCA initiatives as (one time) projects aiming to achieve supply chain objectives using evidence-based problem solving and recognition with a focus on inducing process redesign, tool development or long-term process changes such as automation or continuing decision support.First, the reasoning behind any initiative should be a shortcoming in a supply chain objective (SCO).Either because there is a deficiency in comparison to the theoretic potential or because higher performance is aspired.In the literature, several frameworks of performance dimensions indicating supply chain objectives are proposed [37][38][39][40] without one being unanimously accepted.Several operational metrics reappear in most frameworks we investigated, but the categorization differs tremendously.The following objectives have been elaborated based on a review of these frameworks: cost (SCO1), quality (SCO2), time (SCO3), flexibility (SCO4), sustainability (SCO5), innovativeness (SCO6), customer relationship (SCO7) and supplier reliability (SCO8).
Second, we return to the concept of core capabilities of an analytics competitor: organization, humans, and technology.The organizational aspects can be represented by the analytics objective (AO) of an analytics initiative.In accordance with Manyinka et al. [41] and Holsapple et al. [9] we identified six analytics objectives.Based on evidence-based approaches these include the creation of transparency by democratizing data (AO1), the identification of root causes by experimentation (AO2), the evaluation of business performance and environment (e.g., efficiency or risk assessment) (AO3), the segmentation of populations (including products and services) (AO4), the support and replacement of human decision making (AO5), and the development and innovation of new sources of revenues (e.g., business models, products or services) (AO6).In a given initiative, the fulfilment of one objective can be necessary to pursue a consecutive objective.For example, a transparent business process may be necessary for further analytics approaches leading to supported decision making.
Third, going forward with analytics capabilities, the human involvement (HUM) in a specific initiative can be incorporated by distinguishing the business functions bringing expertise into the initiative.Considering Davenport and Harris [2], Bose [8] and Dietrich et al. [42] we identified several specific roles in analytics initiatives (e.g., several roles from providing access to data to the data management as well as business function from user to sponsor of the initiative) which can both be internal to the organization or externally contracted.We aggregated the roles leading to three groups and six roles: internal analytics expert (HUM1), external analytics expert (HUM2), internal IT expert (HUM3), external IT expert (HUM4), internal business process expert (HUM5), external business process expert (HUM6).As we have seen in our analysis, external and internal expertise is not First, the reasoning behind any initiative should be a shortcoming in a supply chain objective (SCO).Either because there is a deficiency in comparison to the theoretic potential or because higher performance is aspired.In the literature, several frameworks of performance dimensions indicating supply chain objectives are proposed [37][38][39][40] without one being unanimously accepted.Several operational metrics reappear in most frameworks we investigated, but the categorization differs tremendously.The following objectives have been elaborated based on a review of these frameworks: cost (SCO1), quality (SCO2), time (SCO3), flexibility (SCO4), sustainability (SCO5), innovativeness (SCO6), customer relationship (SCO7) and supplier reliability (SCO8).

Supply Chain Objective
Second, we return to the concept of core capabilities of an analytics competitor: organization, humans, and technology.The organizational aspects can be represented by the analytics objective (AO) of an analytics initiative.In accordance with Manyinka et al. [41] and Holsapple et al. [9] we identified six analytics objectives.Based on evidence-based approaches these include the creation of transparency by democratizing data (AO1), the identification of root causes by experimentation (AO2), the evaluation of business performance and environment (e.g., efficiency or risk assessment) (AO3), the segmentation of populations (including products and services) (AO4), the support and replacement of human decision making (AO5), and the development and innovation of new sources of revenues (e.g., business models, products or services) (AO6).In a given initiative, the fulfilment of one objective can be necessary to pursue a consecutive objective.For example, a transparent business process may be necessary for further analytics approaches leading to supported decision making.
Third, going forward with analytics capabilities, the human involvement (HUM) in a specific initiative can be incorporated by distinguishing the business functions bringing expertise into the initiative.Considering Davenport and Harris [2], Bose [8] and Dietrich et al. [42] we identified several specific roles in analytics initiatives (e.g., several roles from providing access to data to the data management as well as business function from user to sponsor of the initiative) which can both be internal to the organization or externally contracted.We aggregated the roles leading to three groups and six roles: internal analytics expert (HUM1), external analytics expert (HUM2), internal IT expert (HUM3), external IT expert (HUM4), internal business process expert (HUM5), external business process expert (HUM6).As we have seen in our analysis, external and internal expertise is not mutually exclusive.Depending on the complexity of the initiative, organizations combine available expertise in various forms to achieve success.
Fourth, for the technological (TEC) aspects and final capabilities the infrastructure can be a major barrier [28].However, Davenport and Harris [2] direct the focus on tools and analytics architecture.They identify small and short initiatives done with spreadsheet software (TEC1) which can be applied by analytical amateurs.Analytical professionals however will either use statistical software (TEC2) for experimental purpose or define and refine specific analytical algorithms (TEC3) building new and often purpose-specific tools.We distinguish the last two, since this algorithm might be bought from a third-party vendor.On the other hand, Davenport and Harris [2] discuss the importance of data storage and access.However, since their initial work, the field has seen significant developments.Opposed to statistical software models gathering data from existing systems or sending it to a third-party to execute the analytics methods, the classical mode is a newly purpose-built data storage (TEC4).Recently, the concept of a non-purpose-built data store to gather data from (TEC5) is emerging following the idea of creating a single storage for analytical purposes to be defined later.This concept is often called "Data Lake" [43].Finally, the concept of virtualization, as prominently known due to cloud computing, allows access to analytical methods and results disconnected from the actual data infrastructure (TEC6), e.g., with mobile devices [44].
Fifth, data management (DAT) for analytical purposes can face serious challenges demanding supplementary effort [33].As explained above, the big data concept represents serious data management challenges, which we have thus incorporated into the evaluation.This includes a high velocity of data (DAT1) being analyzed or collected, a high volume of data (DAT2) to be included in Analytics and a variety of data sources, data structures and data semantics (DAT3).Further, besides internal data (DAT4), managers are supposed to be creative about the inclusion of external data (DAT5) sources [27].
Sixth, reconsidering the works of Hahn and Packowski [23], and Holsapple et al. [9], the type of analytics (TA) can be descriptive (TA1), predictive (TA2), prescriptive (TA3) or several of them at once due to chaining or combination.
As mentioned in the introduction, the initiative's outcome has been neglected in the analysis process.The outcome, which is presented in a percentage or absolute value for savings or improvements in a monetary value, time or quantity is highly dependent on the individual case, organization, and industry.Thus, it was not considered meaningful for the derivation of archetypes.Additionally, this research aims to recognize what organizations aspire, the intention and the consequential execution in a qualitative manner.Quantitative outcomes do not fit this aim.We further omitted firm size and organizational form, since our interest is in the initiatives presented by single projects, which can be a relative small size compared to the size of the organization due to the intention to solve a small problem.
No characteristic described above is mutually exclusive and some will correlate since the presence of one characteristic may likely demand the presence of another.

Methodology
To identify SCA initiative archetypes, we used the machine learning method of clustering.Clustering is a descriptive or explorative data analysis technique which relies on interpretation by the analyst based on insight into the original data [45].This fits MacInnis [24] requirement to use analytical reasoning for facilitating the aspired differentiation.Below, we present the data collection, analysis, and evaluation process.

Data Collection
Since this research considers case studies from organizations, research databases did not provide a sufficient source.Based on the insight we gained from the publications presented above we used key words and synonyms of Analytics ("Data Science", "Business Intelligence", "Big Data" or "Data Mining") as well as Analytics Objective (see Section 2.3) in combination with LSCM ("transport*", "operation management", "deliver*", "value chain", "warehous*", "supplier", "resource planning", "inventory", "material flow", "product handling", "distribut*", "shipping") to conduct an extensive search via the google search engine (with customized search results deactivated).Besides case studies from organizations, we identified several third-party websites, software and solution vendors and organizations applying Analytics, as well as news websites, expert websites, and blogs which we further used for snowball sampling.Finally, we approached organizations for cases.In total, we identified a shortlist of 49 Initiatives with promising information richness to evaluate their characteristics.

Data Analysis
To identify archetypes, we looked at previous research similar to our intent.(L)SCM archetypes aimed at providing managers with understanding of organizational adaptation and performance evaluation have been identified by non-hierarchical clustering on supply chain IT and organizational structure variables [46].The variables were collected via a survey including variables such as B2B e-commerce supply chain integration, ERP applications, operational centralization as well as market and financial performance.Supply chain integration archetypes were investigated to understand the relationship between integration and performance and to provide parsimonious descriptions useful for discussion, research, and pedagogy as well as to reveal insight into the underlying structure [47].Survey-based collection of variables of customer integration, supplier integration, internal integration, business performance and operational performance and the data analysis with hierarchical and subsequent non-hierarchical clustering determined the archetypes.The authors briefly describe five archetypes with three balanced integration archetypes and two customer-leaning integration archetypes in different nuances.LSCM job type archetypes were deviated from collected job descriptions from a major employment website to provide suggestions on how training and professional development should occur [48].Text analysis was used to mechanically code the job descriptions and hierarchical cluster analysis using Ward's method was applied to identify eight archetypes.Concluding, clustering has proven as research method in LSCM to identify archetypes with the method adapted to the individual dataset.Thus, focusing on clustering as the method for our research is supported by previous LSCM research.
We used the 34 characteristics of SCA initiatives discussed above as binary measures to systematically describe the found initiatives.Two researchers coded the cases independently.The cases were coded from the perspective of the analytics result's final user and with the supply chain objective focused on the value creation process.If a case provided inconclusive evidence for a variable, the information was sought from the organization or the case was rejected.Thus, three of the shortlisted case studies were rejected.Both researchers discussed the coding regularly to align the interpretation of the variables.After coding, each difference was discussed and resolved by consensuses.The researchers calculated Cohen's Kappa as 0.65 [49], indicating a substantial agreement on the scale of Landis and Koch [50].
As the collected data is binary, common methods to determine dissimilarities between two objects such as computing the Euclidean or Manhattan distances cannot be employed.To deal with this type of the data, the approach proposed by Kaufman and Rousseeuw [45] building on an adapted version of the similarity coefficient defined by Gower [51] was used to calculate the dissimilarity matrix.All variables were treated as asymmetric binary as they did not represent the presence or absence of a characteristic but the presence and non-presence due to missing evidence for presence [52], thus leading to adaptions of the distance measurement.We used the statistical software 'R' to perform the clustering and its evaluation.
We decided for hierarchical clustering due to the advantage of visually inspecting the agglomerations via a dendrogram.We tested UPGMA, Ward's method, complete linkage and WPGMA and evaluated the dendrograms by outlier influence due to late agglomerations of single observations.The most promising method in our evaluation has been Ward's Method visualized in Figure 2.During the process, we had to omit (DAT_4), since every case relied on internal data.As Hair et al. [53] point out, there is no completely objective way to determine the number of clusters and the final choice remains to the researcher.However, the researcher shall be guided by his research objective.Since we want to create reasonable clusters while keeping them conceptionally knowledgeable and insightful, we limited our considerations to a maximum of eight clusters and decided for the best number of clusters in that range based on several criteria: The change in agglomeration distances between merging clusters peaks at the changes from three to four clusters and from five to six clusters (non-consecutive intersections).Based on the dendrogram, six clusters represent a reasonable cut-off.The Dunn index [54] is maximized for five clusters, while the index remains constant for six and seven clusters.The Silhouette index [55] suggests seven clusters with six clusters as second best and five clusters as third best choice.As shown in Figure 3, the difference between five and six clusters is more pronounced than that between six and seven clusters.We decided for six clusters, since it is the visually most reasonable, once among the best and once the second best considering the indexes.In the visual evaluation, we considered the changes in distance from five to six cluster or six to seven not being consecutive intersections.The resulting six clusters were interpreted by considering the initiatives in each cluster and by comparing the averages of variables among clusters.objective.Since we want to create reasonable clusters while keeping them conceptionally knowledgeable and insightful, we limited our considerations to a maximum of eight clusters and decided for the best number of clusters in that range based on several criteria: The change in agglomeration distances between merging clusters peaks at the changes from three to four clusters and from five to six clusters (non-consecutive intersections).Based on the dendrogram, six clusters represent a reasonable cut-off.The Dunn index [54] is maximized for five clusters, while the index remains constant for six and seven clusters.The Silhouette index [55] suggests seven clusters with six clusters as second best and five clusters as third best choice.As shown in Figure 3, the difference between five and six clusters is more pronounced than that between six and seven clusters.We decided for six clusters, since it is the visually most reasonable, once among the best and once the second best considering the indexes.In the visual evaluation, we considered the changes in distance from five to six cluster or six to seven not being consecutive intersections.The resulting six clusters were interpreted by considering the initiatives in each cluster and by comparing the averages of variables among clusters.

Results and Discussion
In this section, we describe the six found clusters by highlighting characteristics and combinations of characteristics of the initiatives in each cluster in comparison with the initiatives in the others.The findings therefore present the researchers' interpretation.All initiatives within a cluster were then analyzed together to extract commonalities.Clusters have been named to underline their major traits.The results are visualized in Figure 4 with key points for every characteristics category in the order presented in Section 2.3.objective.Since we want to create reasonable clusters while keeping them conceptionally knowledgeable and insightful, we limited our considerations to a maximum of eight clusters and decided for the best number of clusters in that range based on several criteria: The change in agglomeration distances between merging clusters peaks at the changes from three to four clusters and from five to six clusters (non-consecutive intersections).Based on the dendrogram, six clusters represent a reasonable cut-off.The Dunn index [54] is maximized for five clusters, while the index remains constant for six and seven clusters.The Silhouette index [55] suggests seven clusters with six clusters as second best and five clusters as third best choice.As shown in Figure 3, the difference between five and six clusters is more pronounced than that between six and seven clusters.We decided for six clusters, since it is the visually most reasonable, once among the best and once the second best considering the indexes.In the visual evaluation, we considered the changes in distance from five to six cluster or six to seven not being consecutive intersections.The resulting six clusters were interpreted by considering the initiatives in each cluster and by comparing the averages of variables among clusters.

Results and Discussion
In this section, we describe the six found clusters by highlighting characteristics and combinations of characteristics of the initiatives in each cluster in comparison with the initiatives in the others.The findings therefore present the researchers' interpretation.All initiatives within a cluster were then analyzed together to extract commonalities.Clusters have been named to underline their major traits.The results are visualized in Figure 4 with key points for every characteristics category in the order presented in Section 2.3.

Results and Discussion
In this section, we describe the six found clusters by highlighting characteristics and combinations of characteristics of the initiatives in each cluster in comparison with the initiatives in the others.
The findings therefore present the researchers' interpretation.All initiatives within a cluster were then analyzed together to extract commonalities.Clusters have been named to underline their major traits.The results are visualized in Figure 4 with key points for every characteristics category in the order presented in Section 2.3.

Cluster 1-Educating
The initiatives in the Educating cluster focus on gathering data (sources) new to the organization and process, that are used as more advanced input in the decision-making of the LSCM process to improve output but will not lead to process redesign.This cluster contains mainly forecasting initiatives with the specific goal of providing the right amount of goods without storing excess inventory or having stock-outs such that the consumer of goods is served best (e.g., the use of data on weather, income, historical sales, or regional marketing campaigns to determine inventory

Cluster 1-Educating
The initiatives in the Educating cluster focus on gathering data (sources) new to the organization and process, that are used as more advanced input in the decision-making of the LSCM process to improve output but will not lead to process redesign.This cluster contains mainly forecasting initiatives with the specific goal of providing the right amount of goods without storing excess inventory or having stock-outs such that the consumer of goods is served best (e.g., the use of data on weather, income, historical sales, or regional marketing campaigns to determine inventory allocation to individual stores).Thus, the objective is to improve process quality and customer relationship.The tools used in these initiatives are dominantly predictive and aim to produce more precise inputs for decision-making in consecutive processes of inventory allocation.In some initiatives, no evidence on how the consecutive processes are further affected by the tool can be found.The report on one initiative specifically states that this is intended, since employees shall use the output of the model-the prediction-and not question it.The focal organization usually buys a custom-built or customized advanced tool or system extension developed by external Analytics experts, which includes external data sources new to the forecasting organization and combines it with internal data from existing systems to integrate it with these systems.

Cluster 2-Observing
The initiatives in the Observing cluster concentrate heavily on predicting LSCM process deficiencies in the short-term or medium-term future with a newly developed tool observing and monitoring the processes gain reaction time to either prevent the deficiencies or enable counteraction.The prediction is based on process conditions indicating those definitely identified in the initiative.Thus, observing indicates watching with knowing what to pay attention to.The supply chain objectives are mixed but tend to focus on cost reduction and quality improvement.The process is sought to be improved by observing and avoiding identified causes of quality deficiencies but not essentially by changing or redesigning the process.The initiatives in Cluster 2 describe a variety of data experiments to improve process accuracy and quality (e.g., identifying production quality indicators that must be monitored, estimation of product weight based on package weights to sequentially use package weight as quality indicator for correct items, identify influencing factors on punctuality of arrival to adjust plans when factors are present).In a distinct proportion of initiatives, the maintenance process of an asset, machine, or vehicle was under investigation.The actors in the analytics team are mixed from internal and external experts for Analytics and IT.Process experts are usually in house.The software used includes spreadsheets-the cluster contains the only initiative the authors could identify using spreadsheet software (in combination with statistical software)-and statistical software but no specifically designed algorithm.The techniques used are primarily predictive and analytically aimed at anticipating the behavior and evaluating the performance of a process as well as understanding the causes of the process behavior.Thus, they usually exploit patterns in the data opposed to process knowledge as in Alerting initiatives.Data is either stored in purpose-built data storages or gathered from existing systems.In addition, cloud computing was used in some initiatives to provide access to the automatically evaluated processes to facilitate monitoring.External data is barely used, but internal data comes in high volumes and variety to create sophisticated process insight tools.

Cluster 3-Alerting
The initiatives in Cluster 3 use similar approaches as the initiatives of Cluster 2. However, the product of the initiative diverges with Cluster 3 initiatives aiming to produce a support system for process owners.These are supposed to call for attention or alert in certain, especially critical, process conditions or events, which are mostly predefined rather than identified.Critical refers to negative process effects or possible loss of revenue.The on-demand attention contributes to meet various supply chain objectives including cost reduction, quality improvement, flexibility increase or customer relationship improvement.The LSCM process is usually untouched, as opposed to Advancing or Refining initiatives, while the monitoring task of the process is reduced from active checking to passively getting alerted on actions required (e.g., by providing alerts when delivery vehicles do not progress on route as has been estimated which demands modification of routes, informing receivers or parallel deliveries with faster delivery time).To achieve this, the initiatives repeatedly describe teams of internal process experts teaming up with mixed experts in analytics and IT.The need for external assistance in these initiatives may be attributed to the desired output, which is to develop a dedicated software tool in all initiatives in Cluster 3.These tools perform monitoring tasks of a specific process and recommend actions for improvements (e.g., to lower energy consumption, to increase flexibility, to reduce cost, to improve utilization of capacity) as well as request needed actions (e.g., change routes, change active supplier, maintain machines, to adjust prices).Thus, high velocity of data analysis is in focus.The organizations in some cases used the tool to offer new services to their customers.The tools usually need their own purpose-built data storage with virtualization technologies commonly used for improved access.Further, they are likely to include external data sources necessary for risk assessments of suppliers, sources of delays on routes or condition evaluation supported by additional manufacturer provided data.As opposed to Observing initiatives, these initiatives are commonly driven by process knowledge of critical conditions and therefore focus on finding data-driven ways to automate what process owners were actively monitoring before.The initiatives use and combine descriptive and predictive techniques to summarize data for monitoring and extrapolating future conditions.

Cluster 4-Advancing
The initiatives in Cluster 4 focus on advancing a process, or even the organization by developing new business models.Thus, as opposed to Observing initiatives, which optimize reacting on process conditions potentially leading to process quality deprivation, or to Alerting initiatives, in which known conditions require actions, these initiatives usually introduce process changes in the form of adjusted process steps that were identified based on process data made available to create this transparency-the large scale data availability is often seen as major progress and benefit with resulting tool focused process execution.This cluster unites initiatives of organizations with a certain maturity in Analytics.These usually bring internal expertise into the initiative from all relevant areas and only occasionally require external expertise.In particular, the initiatives aim to achieve data availability and transparency and subsequently understand the focal process (e.g., by equipping assets with a variety with sensors and mobile devices, collecting data, and analyzing data from similar assets together to determine the life cycle process of a machine and its components or routines of transport vehicles performing deliveries and consequentially creating a new form of predictive maintenance contract with customers).In this context, "understanding" not only refers to the continuous evaluation of processes, but also to the use of descriptive techniques for causal analysis to identify process parameter settings and combinations causing losses in process quality, and to provide a decision support to counteract these losses.The initiatives integrate data with high velocity, high volume, and high variety.To achieve this, purpose-built software tools are developed.The results of the analysis lead to tools for process monitoring combined with decision support systems.These initiatives further emphasize the collection of data without predefined purpose, with prospective use of these centralized data in future analyses.The tools and select data are regularly made available to customers to create a competitive advantage for their own products.

Cluster 5-Refining
The initiatives in Cluster 5 aim to squeeze the last bit of untapped efficiency out of a system by refining processes with assisting or rather guidance functions for human operatives to reduce costs and save time.Additionally, these initiatives have a strong focus on creating an innovative advantage over competition.To achieve this, prescriptive techniques are used to determine the best course of action-redesigning the process with manifold interactions with the system to guide decision-making by human operatives.This includes the transfer of routing algorithms to pickers and the stops of their carts in distribution centers, augmenting delivery vehicle drivers with routing algorithms dynamically using real-time traffic conditions to re-optimize while the vehicle is already on the road or extending manufacturing processes with real-time quality evaluation to change the production sequence.Thus, as opposed to the archetypes above, the focus diverges from understanding and monitoring a process to continuously adjust (or refine) it.To achieve the aspired goal, a high volume of high velocity data must be analyzed by purpose-built tools.The highly customized tools as well as the systems developed to execute these tools are developed in-house with a combination of internal and external expertise.

Cluster 6-Investigating
The initiatives in cluster 6 are united by the investigation of causes of deficiencies or rather major process flaws.Thus, the objective of these initiatives is to increase process quality and reduce costs by identifying and mitigating the root causes of process flaws or finding proxies to enable monitoring them (e.g., identify shelf replenishment flaws by monitoring check-out patterns in super markets to identify patterns of lost sales or identifying critical sensor signals presenting factors causing quality issues in production processes).The important aspect to distinguish this cluster from the others is the consequences taken when a root cause is found.Handling the root causes in the initiatives could not be done by automating decision making or more sophisticated monitoring.Rather, the cause of process reliability or deficiency must be handled by changes in new product development, major process redesign or changes in materials demanding creativity and engineering design.These initiatives usually combine external Analytics with mixed external and internal IT and processes expertise.Identifying causalities is supposed to start a solution search instead of automating evaluation or continuous decision support, as compared to Observing, and Alerting or Refining initiatives.The results of the initiatives are based on statistical software with a focus on internally available data.Purpose-built data storage systems are created.

Discussion on Archetypes
The clusters presented above present archetypical initiatives of Analytics in LSCM.The identification and interpretation of core characteristics of clusters was conducted to highlight the uniqueness of each archetype and present archetype diversity in intended problem to be solved, execution, techniques, and product.Single initiatives forming the clusters and therefore determining the archetypes differ in some characteristics from the archetype.Thus, new initiatives may be created with differences in single features but with a clearer understanding of archetypical feature combination.In addition, while there is some (expected) overlap of clusters, we consider it as new insight that e.g., the same Analytical objective may be used to pursue different Supply Chain objectives.
Considering the identified archetypes, as well as the characteristics forming the clusters, we observed that the type of Analytics did not dominate.While the Refining archetype consists solely of optimization initiatives, optimization techniques could be found in the Educating and Advising archetypes as well.Predictive techniques can be found in all archetypes, including the Refining archetype, since techniques were usually combined in more sophisticated Analytics initiatives.This holds true for descriptive initiatives as well.While the type of Analytics has been our greatest concern, we additionally conducted an analysis for dominating characteristics by evaluating all characteristics across all archetypes in search for characteristics present in all initiatives forming an archetype but not present in any other archetype.However, we could not find any characteristic fulfilling this condition of dominance.To extend our analysis of critical characteristics, we considered whether the supermajority (two-thirds) of initiatives possessing a certain characteristic are given in any archetype.This condition was defined as weak dominance for this research.This condition was fulfilled by the features SCO8, TEC1 and TEC5.However, these features have two, one, and three observed initiatives possessing the characteristic, respectively.Therefore, we did not consider these features as critical.Concluding, we are confident in our results not being dominated by one single characteristic.
When presenting these results to scholars, a major point of controversy has been, whether the archetypes present levels of Analytics maturity of an organization which we reject after careful consideration due to the following aspects: First, considering the given data, one initiative does not reflect the whole organization but a business unit executing the initiative.This is consistent with research indicating Analytics should follow process maturity [12] or rather additional Analytics Maturity should fit process maturity [11] which is therefore not necessarily leveled across the organization.Second, organizations could execute initiatives of lower maturity since it may still provide benefits and initiatives of higher maturity using external support.Thus, an initiative is not a distinct confirmation of an organizations capabilities.Third, in the initiatives considered, two organizations have been observed twice and one organization three times.The initiatives thus spread across archetypes with the seemingly more mature initiatives either in the same year or earlier.
Fourth, research has pointed out, that the objective of an Analytics initiative is often set without the consideration of the complexity of the Analytics required [25].While the objective guides the initiative, the necessary Analytics maturity may be determined during the execution and not before and thus not influence the initiative.However, we acknowledge that the level of internal expertise involved may indicate the business criticality.
Concerning an initiative perspective, this actually opposes maturity models setting standards for organization-wide implementation for highest maturity [56] or considering strategic initiatives for highest maturity [16].These levels of maturity address the analytics culture of the organization [2] which may influence the spread of initiatives but not dictate the choice of initiatives.
Finally, we learned that, to achieve value with SCA, the solution does not have to be an organization-wide expensive third-party tool.Small models build with R or SPSS and visualized with Tableau can provide significant value already.

Discussion on Overcoming Barriers with Archetypes
This research aspires to provide means to overcome barriers of applying Analytics to LSCM related to a missing understanding.Considering Sanders [28], we chose and summarized several barriers relevant for this research in Section 2.2.
Lack of leadership is indicated to be caused by lack of vision, lack of understanding of the capability, and the lack of understanding how to lead change.The latter, also described as creating a data-oriented culture [57] or data-driven culture [58], is considered a key competency for managers to transform organizations to sophisticated Analytics capabilities and beyond the scope of this research.The lack of vision is addressed by the core concepts of each archetype since they are supposed to guide vision by providing points of reference to individualize, adapt and combine.The lack to understand the capabilities required to apply Analytics is addressed by the characteristics of Analytics initiatives emphasizing structure of and resources needed for executing an initiative.However, it has been suggested that the existence of these capabilities in an organization does not guarantee the ability to bring it to full use [9].
The barriers of lacking objectives are addressed by the generic objectives presented by the two objective characteristics categories, and by the specific objectives provided in the archetype descriptions.This highlights to managers the necessity of defining an objective as compared to the "poking" for correlation as described by the notion of analysis paralysis.Defining an objective which Analytics should answer is a valuable starting point [59].The archetypes further emphasize that Analytics initiatives should not bedriven by the latest and most innovative technology but by technology fitting the purpose of the identified objectives.
Further, as evident from the discussion above, with an orientation on objective-driven SCA with subsequent choice of data, managers should not drown in data.The archetypes further present the opportunity of relying on external guidance for choosing the necessary data.In addition, it is indicated that using non-Big Data can still achieve benefits.This is further underlined by process models for Analytics initiatives recommending data collection to be a later step in the project (e.g., [31,57]).Even while drowning in data, having the right data to successfully execute the Analytics initiative is not guaranteed.

Conclusions
In our research, we investigated how SCA initiatives can be distinguished.Literature suggests reluctance of LSCM to invest in Analytics initiatives caused among other reasons by managers missing ideas in how to approach SCA.With our research, we address this shortage by providing a distinction of initiatives providing knowledge to managers about typical approaches to use SCA to gain business value.Based on the patterns emerging form a cluster analysis of 46 SCA initiatives we propose six archetypes that show considerable differences in how organizations deploy SCA.In the analysis, the problem to be solved, execution, techniques, and resulting Analytics solution of the initiative have been considered.In detail, we examined characteristics necessary to execute an SCA initiative and therefore display areas that must be considered by managers designing new initiatives.The characteristics are aggregated into the following groups:

•
Supply chain objective that shall be addressed which represents the problem or deficiency in the LSCM process; Regarding the groups of characteristics above, our findings support considerable differences in initiative archetypes.The patterns identified allow us to answer the research question: SCA initiatives can be distinguished in the six clusters which are described regarding the characteristics in Section 4 as well as LSCM process centric as follows: 1.
Educating: The LSCM process remains as existing but will be enhanced with new data (sources) information as process input to improve decisions to be made during the process resulting in enhanced LSCM process output quality and customer orientation.This typically emerges as an improved tool used in the process such as a new forecasting model in a product allocation process or new forecast model for a risk evaluation process.

2.
Observing: The LSCM process is extensively investigated for conditions that indicate process deficiencies or issues in the short-term or medium-term future with a resulting tool to monitor the process based on the newly gained insight.The knowledge about the conditions improves process quality and costs due to earlier reaction.Examples include detection of engine vibration patterns enabling maintenance planning of vehicles such that a repair shop is the final stop of a route on a suitable point in time instead of random breakdown far away from access to maintenance, or detection of weather patterns resulting in traffic and road conditions demanding changing of routes.However, identified conditions are indications and leave room for human decision making.

3.
Alerting: LSCM process owners are provided with alerts on critical conditions and events that immediately demand reactions.The conditions are usually known by process owners without the need of analytical identification and certain in their negative impact on the process demanding actions.Alerting initiatives' central task is making the necessary data available to automate the alert as opposed to repeated human check-up actions.Examples include alerts on closed roads for vehicle routing or automated recommendations of price changes and acceptance of shipments for cargo airlines in close to departure time-windows.Here again, the LSCM process is typically supported but not altered.

4.
Advancing: The LSCM processes and business models will be advanced by enabling changes due to insight made available with intense data collection and analysis.Large scale data collection is central to the initiative, using sensors and mobile devices to create data-availability-based transparency and evaluation of LSCM process steps.The insight is used to improve process quality by changing process steps under incorporation of the insight and creating analytics driven innovations replacing process steps as well as making insight available to interested third parties as business model innovation.Examples are machine profiles allowing determination of accurate predictive maintenance processes which can be sold by the machine manufacturer to the machine user, or driver profiles to create new monitoring steps to reduce idle time.These initiatives differ from observing and alerting by extensiveness of data collection and analysis typically demanding big data technologies, and range of the resulting tool, which changes the process to become tool and thus data focused as opposed to a minor process support.

5.
Refining: The LSCM processes are changed to incorporating faster, broader, and more frequent guidance on actions and decision support.Instead of optimized plans that are executed, the objective of these initiatives is to optimize plans during execution dynamically based on data about current events and conditions.Examples are dynamic changes of routes of vehicles already on the road, or dynamic changes of picker routes in distribution centers already picking.The LSCM process is changed due to extensive focus on guidance tools guidance during process execution.6.
Investigating: The LSCM process (and asset) deficiencies and issues are investigated for their causes to enable the solution search for design changes to the process.These changes are supposed to create new processes with improved costs and quality over the process under investigation.As opposed to issues described in advancing or refining, process changes such as automation or data-driven tools for guidance will not create control over the process issues addressed in these initiatives.Thus, creativity and engineering design is required.Examples include the investigation of occurrence of empty shelf space in retail stores to redesign replenishment processes of products or the investigation of process environment factors in production lines leading to quality issues that must be avoided.

Theoretical Contribution
Our research contributes to LSCM research with a focus on SCA and the practical application of SCA, an area that has been demanded to be investigated by several researchers in LSCM [14,15,28].The identified archetypes provide an empirically developed taxonomy.Further, they give insight into the underlying structure of Analytics initiatives used in LSCM-why and how they are applied.The research decomposes SCA initiatives in important distinct parts, which can structure future research.Thereby, this research specifically addresses characteristics influencing Analytics initiative and is not limited to distinction by software [23] or LSCM process [10].
The proposed archetypes seek to guide discussions, research and training of students becoming managers enabled to use SCA.The discussion aspired by the authors should address how to enable organizations to create initiatives beyond the presented contemporary archetypes with more sophisticated supply chain and Analytics objectives, rather than conducting single case studies or literature reviews on the competitive impact without empirical evidence.Our research provides a framework supporting the investigation of the effects of different types of Analytics initiatives and helps researchers working with data models and quantitative case studies to orientate themselves in the bigger picture of their research.This framework further allows to investigate the implications of various kinds of SCA initiatives on performance, barriers as well as the efficiency of the initiatives based on the archetype of the initiative.Finally, this research gives a two-dimensional picture to introduce students to this field and ease the process of understanding important factors and possibilities by the proposed first dimension of archetypes to understand what companies do and the second dimension of characteristics of SCA initiatives to understand what aspects to consider when constructing an initiative.Thus, it enables researchers and students to introduce their LSCM knowledge into Analytics initiatives and provide considerable value that is required for successful initiatives [14,15].This research further addresses the gap between theory and organizational activities highlighted by several scholars, especially in management science [60,61].Our archetypes map the activities of organizations and provide templates for organizations and scholars in the field to understand what drives organizations to their activities.

Managerial Contribution
This research copes with the managerial barriers related to missing insight into the application of SCA.By describing archetypes of this application, we give managers directions for future SCA initiatives based on their initial business situation, available means, and objectives.Presenting the results to experts in Analytics, the archetypes were well received with the remark that managers may lack creativity of how to address business problems with Analytics which could be supported with the results of this research.In this regard, managers may combine archetypical approaches to create new initiatives or explore initiatives with supply chain or analytical objectives rarely observed.
Considering the barriers discussed in Section 4.8, our research presents how the application of SCA creates value for an organization and how decisions are made based on SCA.Managers should be enabled to decide which of the overwhelming opportunities provided by SCA to take and which to postpone or reject with the primary objective of providing value to the organization.Naturally, this requires the creativity to design new initiatives.
The research further provides a framework for managers to understand the key components to build an SCA initiative.First and foremost, an initiative must address existing problems-meaning any disparity between objective state and actual state-for the LSCM Part as well as they require an analytical objective to address the LSCM problem.Further characteristics display fields that must be developed and improved over time to design more complex initiatives, even in the same archetype.For the Human category, that includes building skills supporting the execution of initiative as hard skills in Analytics as well as communication skills to transfer thoughts, ideas, and experience between the different experts.In the technology category, this include investments in easier data exchange, faster analysis, and calculation as well as more-powerful analytical tools.In the data category, this includes broader data collection and higher standards for data quality.
Our research further develops a vocabulary to communicate managers' objectives and vision while highlighting small but crucial differences.The archetypes are imagined as a menu of options a manager may use to choose specific or combined items.We intend the archetypes to guide his initiative design process, as opposed to having an infinity of options that quickly becomes overwhelming.Therefore, besides providing directions, the archetypes also serve as validation of the fit of characteristics of the initiative and thus, its practicality.This enables managers to pinpoint what they aspire and communicate it directly and properly.
The other way around, the archetypes can be stimulation for two additional types of managers keen to use Analytics in LSCM.First, Managers that achieved some routine in initiatives of a specific archetype may get stuck in that archetype and repeat it for every new use case which eventually leads to decreasing marginal value from that kind of initiative.Second, the managers are supposed to "make more from their data"-a type that is not very rare from our personal experience.Both types of managers could, using the archetypes, identify promising problems to address and subsequently search for interested users or rather "problem owners".The first type obviously benefits over the second from knowing eventual problem owners from her previous projects she could address again with another beneficial initiative.

Limitations
Due to the various sources of the initiatives considered, their descriptions are provided in various levels of detail regarding the characteristics used to evaluate them.With 46 cases, the amount of considered cases is low.Additionally, the observed cases only represent successes since these are more likely to be published in any form.Unsuccessful cases to use Analytics in LSCM could not be identified.Further, the cases have been collected in a procedure which is hard to recreate.This is due to the lack of a public database for such case studies, especially considering the amount of studies needed to conduct a meaningful cluster analysis.Since databases for research (e.g., Scopus, Web of Science or EBSCO) did not yield relevant results, we were reliant on an open search platform.The search was suspended when a reasonable amount of time (16 h/2 workdays) for cases did not yield any new results.However, the possibilities to collect the data for this research were rather limited.
Furthermore, considering the data analysis, the decision about the number of clusters and thus the number and structure of the identified archetypes depends on several vague factors and cannot be made objectively.The clusters are created based on the researchers' interpretations and judgement.We presented the results to researchers in LSCM and experts in Analytics, which assessed the clusters as reasonable.

Future Research
This study takes a step towards understanding the inner structure of a growing field of research, which should not be investigated as a single entity to generalize use, effects, and benefits anymore.Thus, future research may investigate the effects of SCA initiatives distinguished by archetype to create more sophisticated insight.The archetypes may also be correlated to Analytics maturity or the growth-share matrix to identify easy-to-start archetypical initiatives for organizations with low maturity and identify factors of successful initiatives with the potential to create competitive advantage.Additionally, since we consider this research to be contemporary, we encourage to repeat this research in five to ten years.

Figure 1 .
Figure 1.Characteristics of a Supply Chain Analytics initiative.

Figure 1 .
Figure 1.Characteristics of a Supply Chain Analytics initiative.

Figure 2 .
Figure 2. Dendrogram of Cluster Analysis with Ward's method.

Figure 2 .
Figure 2. Dendrogram of Cluster Analysis with Ward's method.

Figure 2 .
Figure 2. Dendrogram of Cluster Analysis with Ward's method.

Figure 4 .
Figure 4. Proposed Supply Chain Analytics archetypes (no chronology or sequence intended).

•
Analytics objective, which is addressing how data and Analytics are supposed to support, effect, or change the LSCM process;•Human expertise in areas relevant to the initiative as Analytics, IT and the LSCM process (and how it is sourced);